River is building a long-lasting, non-penetrating brain computer interface to restore and augment the human experience.
The device sits epidurally, so it reads high resolution signals without damaging the brain. It is designed to last thirty years or more, which is what someone with aphasia or a spinal cord injury needs from a device they will keep for the rest of their life. River works with the University of Melbourne, Stanford University and Maastricht University. Its first clinical study is in preparation.
Most brain computer interfaces make you choose between a device that reads real thought and a device that leaves the brain undamaged. Oseh Mathias argues that going epidural gets you both, and that the gap between thought and machine is a hardware problem rather than a model problem: "We do not want visible devices. No electrodes all over the face. No headsets. We want true thought. That means inner speech."
Oseh Mathias had already shipped in both of the fields this problem sits between. He trained speech models from the AudioLM, SoundStorm and SpearTTS papers at Superfeel. He built SpeechFit, which families in 34 countries use to treat stuttering and work on fluency. He also ran the commercial and regulatory launch of Tecentriq for Roche across Australia and New Zealand, and a hardware founder who has taken a medical product through a regulator is rare. He then built a team around the same bet: David Eagleman advising, Christian Herff on science, Sam John on preclinical work, with ClearPoint Neuro and Neo-Bionica building alongside them.